复分形网络业务的离散小波谱的特征属性及其参数估计

Discrete Wavelet Spectrum’s Characherization and Its Parameter Estimating for Multifractal Network Traffic

研究发现,高速因特网的合成业务流具有复分形特性.研究统计过程的复分形结构及其特征属性是研究网络业务全尺度下的行为的重要内容.文中基于离散小波变换给出了一种小波谱函数的定义,它可以表征复分形过程的所有阶的统计特性.文中给出了谱函数存在性的证明,并给出了相应的特征属性和条件.最后,基于谱函数的定义,文中给出了小波谱函数的估计方法,并针对实际网络数据进行了小波谱分析.

Recent studies on rapid Internet traffic have shown that aggregate traffic is multipfractal. To characterize the properties and construction of multipfractal products of stochastic processes is important for capturing the traffic behavior at all time scales. This paper defines a spectrum function, obtained from a discrete wavelet transform （DWT）, to represent all-order statistic properties of multifractal process. The existence of the spectrum is given as well as its characterization properties and the corresponding conditions. Based on these, this paper gives a method of spectrum estimating and analyzes the real network data using it at last.